xref: /linux/net/l2tp/l2tp_core.c (revision f79e4d5f92a129a1159c973735007d4ddc8541f3)
1 /*
2  * L2TP core.
3  *
4  * Copyright (c) 2008,2009,2010 Katalix Systems Ltd
5  *
6  * This file contains some code of the original L2TPv2 pppol2tp
7  * driver, which has the following copyright:
8  *
9  * Authors:	Martijn van Oosterhout <kleptog@svana.org>
10  *		James Chapman (jchapman@katalix.com)
11  * Contributors:
12  *		Michal Ostrowski <mostrows@speakeasy.net>
13  *		Arnaldo Carvalho de Melo <acme@xconectiva.com.br>
14  *		David S. Miller (davem@redhat.com)
15  *
16  * This program is free software; you can redistribute it and/or modify
17  * it under the terms of the GNU General Public License version 2 as
18  * published by the Free Software Foundation.
19  */
20 
21 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
22 
23 #include <linux/module.h>
24 #include <linux/string.h>
25 #include <linux/list.h>
26 #include <linux/rculist.h>
27 #include <linux/uaccess.h>
28 
29 #include <linux/kernel.h>
30 #include <linux/spinlock.h>
31 #include <linux/kthread.h>
32 #include <linux/sched.h>
33 #include <linux/slab.h>
34 #include <linux/errno.h>
35 #include <linux/jiffies.h>
36 
37 #include <linux/netdevice.h>
38 #include <linux/net.h>
39 #include <linux/inetdevice.h>
40 #include <linux/skbuff.h>
41 #include <linux/init.h>
42 #include <linux/in.h>
43 #include <linux/ip.h>
44 #include <linux/udp.h>
45 #include <linux/l2tp.h>
46 #include <linux/hash.h>
47 #include <linux/sort.h>
48 #include <linux/file.h>
49 #include <linux/nsproxy.h>
50 #include <net/net_namespace.h>
51 #include <net/netns/generic.h>
52 #include <net/dst.h>
53 #include <net/ip.h>
54 #include <net/udp.h>
55 #include <net/udp_tunnel.h>
56 #include <net/inet_common.h>
57 #include <net/xfrm.h>
58 #include <net/protocol.h>
59 #include <net/inet6_connection_sock.h>
60 #include <net/inet_ecn.h>
61 #include <net/ip6_route.h>
62 #include <net/ip6_checksum.h>
63 
64 #include <asm/byteorder.h>
65 #include <linux/atomic.h>
66 
67 #include "l2tp_core.h"
68 
69 #define L2TP_DRV_VERSION	"V2.0"
70 
71 /* L2TP header constants */
72 #define L2TP_HDRFLAG_T	   0x8000
73 #define L2TP_HDRFLAG_L	   0x4000
74 #define L2TP_HDRFLAG_S	   0x0800
75 #define L2TP_HDRFLAG_O	   0x0200
76 #define L2TP_HDRFLAG_P	   0x0100
77 
78 #define L2TP_HDR_VER_MASK  0x000F
79 #define L2TP_HDR_VER_2	   0x0002
80 #define L2TP_HDR_VER_3	   0x0003
81 
82 /* L2TPv3 default L2-specific sublayer */
83 #define L2TP_SLFLAG_S	   0x40000000
84 #define L2TP_SL_SEQ_MASK   0x00ffffff
85 
86 #define L2TP_HDR_SIZE_SEQ		10
87 #define L2TP_HDR_SIZE_NOSEQ		6
88 
89 /* Default trace flags */
90 #define L2TP_DEFAULT_DEBUG_FLAGS	0
91 
92 /* Private data stored for received packets in the skb.
93  */
94 struct l2tp_skb_cb {
95 	u32			ns;
96 	u16			has_seq;
97 	u16			length;
98 	unsigned long		expires;
99 };
100 
101 #define L2TP_SKB_CB(skb)	((struct l2tp_skb_cb *) &skb->cb[sizeof(struct inet_skb_parm)])
102 
103 static struct workqueue_struct *l2tp_wq;
104 
105 /* per-net private data for this module */
106 static unsigned int l2tp_net_id;
107 struct l2tp_net {
108 	struct list_head l2tp_tunnel_list;
109 	spinlock_t l2tp_tunnel_list_lock;
110 	struct hlist_head l2tp_session_hlist[L2TP_HASH_SIZE_2];
111 	spinlock_t l2tp_session_hlist_lock;
112 };
113 
114 #if IS_ENABLED(CONFIG_IPV6)
115 static bool l2tp_sk_is_v6(struct sock *sk)
116 {
117 	return sk->sk_family == PF_INET6 &&
118 	       !ipv6_addr_v4mapped(&sk->sk_v6_daddr);
119 }
120 #endif
121 
122 static inline struct l2tp_tunnel *l2tp_tunnel(struct sock *sk)
123 {
124 	return sk->sk_user_data;
125 }
126 
127 static inline struct l2tp_net *l2tp_pernet(const struct net *net)
128 {
129 	BUG_ON(!net);
130 
131 	return net_generic(net, l2tp_net_id);
132 }
133 
134 /* Session hash global list for L2TPv3.
135  * The session_id SHOULD be random according to RFC3931, but several
136  * L2TP implementations use incrementing session_ids.  So we do a real
137  * hash on the session_id, rather than a simple bitmask.
138  */
139 static inline struct hlist_head *
140 l2tp_session_id_hash_2(struct l2tp_net *pn, u32 session_id)
141 {
142 	return &pn->l2tp_session_hlist[hash_32(session_id, L2TP_HASH_BITS_2)];
143 
144 }
145 
146 /* Session hash list.
147  * The session_id SHOULD be random according to RFC2661, but several
148  * L2TP implementations (Cisco and Microsoft) use incrementing
149  * session_ids.  So we do a real hash on the session_id, rather than a
150  * simple bitmask.
151  */
152 static inline struct hlist_head *
153 l2tp_session_id_hash(struct l2tp_tunnel *tunnel, u32 session_id)
154 {
155 	return &tunnel->session_hlist[hash_32(session_id, L2TP_HASH_BITS)];
156 }
157 
158 void l2tp_tunnel_free(struct l2tp_tunnel *tunnel)
159 {
160 	sock_put(tunnel->sock);
161 	/* the tunnel is freed in the socket destructor */
162 }
163 EXPORT_SYMBOL(l2tp_tunnel_free);
164 
165 /* Lookup a tunnel. A new reference is held on the returned tunnel. */
166 struct l2tp_tunnel *l2tp_tunnel_get(const struct net *net, u32 tunnel_id)
167 {
168 	const struct l2tp_net *pn = l2tp_pernet(net);
169 	struct l2tp_tunnel *tunnel;
170 
171 	rcu_read_lock_bh();
172 	list_for_each_entry_rcu(tunnel, &pn->l2tp_tunnel_list, list) {
173 		if (tunnel->tunnel_id == tunnel_id) {
174 			l2tp_tunnel_inc_refcount(tunnel);
175 			rcu_read_unlock_bh();
176 
177 			return tunnel;
178 		}
179 	}
180 	rcu_read_unlock_bh();
181 
182 	return NULL;
183 }
184 EXPORT_SYMBOL_GPL(l2tp_tunnel_get);
185 
186 struct l2tp_tunnel *l2tp_tunnel_get_nth(const struct net *net, int nth)
187 {
188 	const struct l2tp_net *pn = l2tp_pernet(net);
189 	struct l2tp_tunnel *tunnel;
190 	int count = 0;
191 
192 	rcu_read_lock_bh();
193 	list_for_each_entry_rcu(tunnel, &pn->l2tp_tunnel_list, list) {
194 		if (++count > nth) {
195 			l2tp_tunnel_inc_refcount(tunnel);
196 			rcu_read_unlock_bh();
197 			return tunnel;
198 		}
199 	}
200 	rcu_read_unlock_bh();
201 
202 	return NULL;
203 }
204 EXPORT_SYMBOL_GPL(l2tp_tunnel_get_nth);
205 
206 /* Lookup a session. A new reference is held on the returned session. */
207 struct l2tp_session *l2tp_session_get(const struct net *net,
208 				      struct l2tp_tunnel *tunnel,
209 				      u32 session_id)
210 {
211 	struct hlist_head *session_list;
212 	struct l2tp_session *session;
213 
214 	if (!tunnel) {
215 		struct l2tp_net *pn = l2tp_pernet(net);
216 
217 		session_list = l2tp_session_id_hash_2(pn, session_id);
218 
219 		rcu_read_lock_bh();
220 		hlist_for_each_entry_rcu(session, session_list, global_hlist) {
221 			if (session->session_id == session_id) {
222 				l2tp_session_inc_refcount(session);
223 				rcu_read_unlock_bh();
224 
225 				return session;
226 			}
227 		}
228 		rcu_read_unlock_bh();
229 
230 		return NULL;
231 	}
232 
233 	session_list = l2tp_session_id_hash(tunnel, session_id);
234 	read_lock_bh(&tunnel->hlist_lock);
235 	hlist_for_each_entry(session, session_list, hlist) {
236 		if (session->session_id == session_id) {
237 			l2tp_session_inc_refcount(session);
238 			read_unlock_bh(&tunnel->hlist_lock);
239 
240 			return session;
241 		}
242 	}
243 	read_unlock_bh(&tunnel->hlist_lock);
244 
245 	return NULL;
246 }
247 EXPORT_SYMBOL_GPL(l2tp_session_get);
248 
249 struct l2tp_session *l2tp_session_get_nth(struct l2tp_tunnel *tunnel, int nth)
250 {
251 	int hash;
252 	struct l2tp_session *session;
253 	int count = 0;
254 
255 	read_lock_bh(&tunnel->hlist_lock);
256 	for (hash = 0; hash < L2TP_HASH_SIZE; hash++) {
257 		hlist_for_each_entry(session, &tunnel->session_hlist[hash], hlist) {
258 			if (++count > nth) {
259 				l2tp_session_inc_refcount(session);
260 				read_unlock_bh(&tunnel->hlist_lock);
261 				return session;
262 			}
263 		}
264 	}
265 
266 	read_unlock_bh(&tunnel->hlist_lock);
267 
268 	return NULL;
269 }
270 EXPORT_SYMBOL_GPL(l2tp_session_get_nth);
271 
272 /* Lookup a session by interface name.
273  * This is very inefficient but is only used by management interfaces.
274  */
275 struct l2tp_session *l2tp_session_get_by_ifname(const struct net *net,
276 						const char *ifname)
277 {
278 	struct l2tp_net *pn = l2tp_pernet(net);
279 	int hash;
280 	struct l2tp_session *session;
281 
282 	rcu_read_lock_bh();
283 	for (hash = 0; hash < L2TP_HASH_SIZE_2; hash++) {
284 		hlist_for_each_entry_rcu(session, &pn->l2tp_session_hlist[hash], global_hlist) {
285 			if (!strcmp(session->ifname, ifname)) {
286 				l2tp_session_inc_refcount(session);
287 				rcu_read_unlock_bh();
288 
289 				return session;
290 			}
291 		}
292 	}
293 
294 	rcu_read_unlock_bh();
295 
296 	return NULL;
297 }
298 EXPORT_SYMBOL_GPL(l2tp_session_get_by_ifname);
299 
300 int l2tp_session_register(struct l2tp_session *session,
301 			  struct l2tp_tunnel *tunnel)
302 {
303 	struct l2tp_session *session_walk;
304 	struct hlist_head *g_head;
305 	struct hlist_head *head;
306 	struct l2tp_net *pn;
307 	int err;
308 
309 	head = l2tp_session_id_hash(tunnel, session->session_id);
310 
311 	write_lock_bh(&tunnel->hlist_lock);
312 	if (!tunnel->acpt_newsess) {
313 		err = -ENODEV;
314 		goto err_tlock;
315 	}
316 
317 	hlist_for_each_entry(session_walk, head, hlist)
318 		if (session_walk->session_id == session->session_id) {
319 			err = -EEXIST;
320 			goto err_tlock;
321 		}
322 
323 	if (tunnel->version == L2TP_HDR_VER_3) {
324 		pn = l2tp_pernet(tunnel->l2tp_net);
325 		g_head = l2tp_session_id_hash_2(l2tp_pernet(tunnel->l2tp_net),
326 						session->session_id);
327 
328 		spin_lock_bh(&pn->l2tp_session_hlist_lock);
329 
330 		hlist_for_each_entry(session_walk, g_head, global_hlist)
331 			if (session_walk->session_id == session->session_id) {
332 				err = -EEXIST;
333 				goto err_tlock_pnlock;
334 			}
335 
336 		l2tp_tunnel_inc_refcount(tunnel);
337 		hlist_add_head_rcu(&session->global_hlist, g_head);
338 
339 		spin_unlock_bh(&pn->l2tp_session_hlist_lock);
340 	} else {
341 		l2tp_tunnel_inc_refcount(tunnel);
342 	}
343 
344 	hlist_add_head(&session->hlist, head);
345 	write_unlock_bh(&tunnel->hlist_lock);
346 
347 	return 0;
348 
349 err_tlock_pnlock:
350 	spin_unlock_bh(&pn->l2tp_session_hlist_lock);
351 err_tlock:
352 	write_unlock_bh(&tunnel->hlist_lock);
353 
354 	return err;
355 }
356 EXPORT_SYMBOL_GPL(l2tp_session_register);
357 
358 /*****************************************************************************
359  * Receive data handling
360  *****************************************************************************/
361 
362 /* Queue a skb in order. We come here only if the skb has an L2TP sequence
363  * number.
364  */
365 static void l2tp_recv_queue_skb(struct l2tp_session *session, struct sk_buff *skb)
366 {
367 	struct sk_buff *skbp;
368 	struct sk_buff *tmp;
369 	u32 ns = L2TP_SKB_CB(skb)->ns;
370 
371 	spin_lock_bh(&session->reorder_q.lock);
372 	skb_queue_walk_safe(&session->reorder_q, skbp, tmp) {
373 		if (L2TP_SKB_CB(skbp)->ns > ns) {
374 			__skb_queue_before(&session->reorder_q, skbp, skb);
375 			l2tp_dbg(session, L2TP_MSG_SEQ,
376 				 "%s: pkt %hu, inserted before %hu, reorder_q len=%d\n",
377 				 session->name, ns, L2TP_SKB_CB(skbp)->ns,
378 				 skb_queue_len(&session->reorder_q));
379 			atomic_long_inc(&session->stats.rx_oos_packets);
380 			goto out;
381 		}
382 	}
383 
384 	__skb_queue_tail(&session->reorder_q, skb);
385 
386 out:
387 	spin_unlock_bh(&session->reorder_q.lock);
388 }
389 
390 /* Dequeue a single skb.
391  */
392 static void l2tp_recv_dequeue_skb(struct l2tp_session *session, struct sk_buff *skb)
393 {
394 	struct l2tp_tunnel *tunnel = session->tunnel;
395 	int length = L2TP_SKB_CB(skb)->length;
396 
397 	/* We're about to requeue the skb, so return resources
398 	 * to its current owner (a socket receive buffer).
399 	 */
400 	skb_orphan(skb);
401 
402 	atomic_long_inc(&tunnel->stats.rx_packets);
403 	atomic_long_add(length, &tunnel->stats.rx_bytes);
404 	atomic_long_inc(&session->stats.rx_packets);
405 	atomic_long_add(length, &session->stats.rx_bytes);
406 
407 	if (L2TP_SKB_CB(skb)->has_seq) {
408 		/* Bump our Nr */
409 		session->nr++;
410 		session->nr &= session->nr_max;
411 
412 		l2tp_dbg(session, L2TP_MSG_SEQ, "%s: updated nr to %hu\n",
413 			 session->name, session->nr);
414 	}
415 
416 	/* call private receive handler */
417 	if (session->recv_skb != NULL)
418 		(*session->recv_skb)(session, skb, L2TP_SKB_CB(skb)->length);
419 	else
420 		kfree_skb(skb);
421 }
422 
423 /* Dequeue skbs from the session's reorder_q, subject to packet order.
424  * Skbs that have been in the queue for too long are simply discarded.
425  */
426 static void l2tp_recv_dequeue(struct l2tp_session *session)
427 {
428 	struct sk_buff *skb;
429 	struct sk_buff *tmp;
430 
431 	/* If the pkt at the head of the queue has the nr that we
432 	 * expect to send up next, dequeue it and any other
433 	 * in-sequence packets behind it.
434 	 */
435 start:
436 	spin_lock_bh(&session->reorder_q.lock);
437 	skb_queue_walk_safe(&session->reorder_q, skb, tmp) {
438 		if (time_after(jiffies, L2TP_SKB_CB(skb)->expires)) {
439 			atomic_long_inc(&session->stats.rx_seq_discards);
440 			atomic_long_inc(&session->stats.rx_errors);
441 			l2tp_dbg(session, L2TP_MSG_SEQ,
442 				 "%s: oos pkt %u len %d discarded (too old), waiting for %u, reorder_q_len=%d\n",
443 				 session->name, L2TP_SKB_CB(skb)->ns,
444 				 L2TP_SKB_CB(skb)->length, session->nr,
445 				 skb_queue_len(&session->reorder_q));
446 			session->reorder_skip = 1;
447 			__skb_unlink(skb, &session->reorder_q);
448 			kfree_skb(skb);
449 			continue;
450 		}
451 
452 		if (L2TP_SKB_CB(skb)->has_seq) {
453 			if (session->reorder_skip) {
454 				l2tp_dbg(session, L2TP_MSG_SEQ,
455 					 "%s: advancing nr to next pkt: %u -> %u",
456 					 session->name, session->nr,
457 					 L2TP_SKB_CB(skb)->ns);
458 				session->reorder_skip = 0;
459 				session->nr = L2TP_SKB_CB(skb)->ns;
460 			}
461 			if (L2TP_SKB_CB(skb)->ns != session->nr) {
462 				l2tp_dbg(session, L2TP_MSG_SEQ,
463 					 "%s: holding oos pkt %u len %d, waiting for %u, reorder_q_len=%d\n",
464 					 session->name, L2TP_SKB_CB(skb)->ns,
465 					 L2TP_SKB_CB(skb)->length, session->nr,
466 					 skb_queue_len(&session->reorder_q));
467 				goto out;
468 			}
469 		}
470 		__skb_unlink(skb, &session->reorder_q);
471 
472 		/* Process the skb. We release the queue lock while we
473 		 * do so to let other contexts process the queue.
474 		 */
475 		spin_unlock_bh(&session->reorder_q.lock);
476 		l2tp_recv_dequeue_skb(session, skb);
477 		goto start;
478 	}
479 
480 out:
481 	spin_unlock_bh(&session->reorder_q.lock);
482 }
483 
484 static int l2tp_seq_check_rx_window(struct l2tp_session *session, u32 nr)
485 {
486 	u32 nws;
487 
488 	if (nr >= session->nr)
489 		nws = nr - session->nr;
490 	else
491 		nws = (session->nr_max + 1) - (session->nr - nr);
492 
493 	return nws < session->nr_window_size;
494 }
495 
496 /* If packet has sequence numbers, queue it if acceptable. Returns 0 if
497  * acceptable, else non-zero.
498  */
499 static int l2tp_recv_data_seq(struct l2tp_session *session, struct sk_buff *skb)
500 {
501 	if (!l2tp_seq_check_rx_window(session, L2TP_SKB_CB(skb)->ns)) {
502 		/* Packet sequence number is outside allowed window.
503 		 * Discard it.
504 		 */
505 		l2tp_dbg(session, L2TP_MSG_SEQ,
506 			 "%s: pkt %u len %d discarded, outside window, nr=%u\n",
507 			 session->name, L2TP_SKB_CB(skb)->ns,
508 			 L2TP_SKB_CB(skb)->length, session->nr);
509 		goto discard;
510 	}
511 
512 	if (session->reorder_timeout != 0) {
513 		/* Packet reordering enabled. Add skb to session's
514 		 * reorder queue, in order of ns.
515 		 */
516 		l2tp_recv_queue_skb(session, skb);
517 		goto out;
518 	}
519 
520 	/* Packet reordering disabled. Discard out-of-sequence packets, while
521 	 * tracking the number if in-sequence packets after the first OOS packet
522 	 * is seen. After nr_oos_count_max in-sequence packets, reset the
523 	 * sequence number to re-enable packet reception.
524 	 */
525 	if (L2TP_SKB_CB(skb)->ns == session->nr) {
526 		skb_queue_tail(&session->reorder_q, skb);
527 	} else {
528 		u32 nr_oos = L2TP_SKB_CB(skb)->ns;
529 		u32 nr_next = (session->nr_oos + 1) & session->nr_max;
530 
531 		if (nr_oos == nr_next)
532 			session->nr_oos_count++;
533 		else
534 			session->nr_oos_count = 0;
535 
536 		session->nr_oos = nr_oos;
537 		if (session->nr_oos_count > session->nr_oos_count_max) {
538 			session->reorder_skip = 1;
539 			l2tp_dbg(session, L2TP_MSG_SEQ,
540 				 "%s: %d oos packets received. Resetting sequence numbers\n",
541 				 session->name, session->nr_oos_count);
542 		}
543 		if (!session->reorder_skip) {
544 			atomic_long_inc(&session->stats.rx_seq_discards);
545 			l2tp_dbg(session, L2TP_MSG_SEQ,
546 				 "%s: oos pkt %u len %d discarded, waiting for %u, reorder_q_len=%d\n",
547 				 session->name, L2TP_SKB_CB(skb)->ns,
548 				 L2TP_SKB_CB(skb)->length, session->nr,
549 				 skb_queue_len(&session->reorder_q));
550 			goto discard;
551 		}
552 		skb_queue_tail(&session->reorder_q, skb);
553 	}
554 
555 out:
556 	return 0;
557 
558 discard:
559 	return 1;
560 }
561 
562 /* Do receive processing of L2TP data frames. We handle both L2TPv2
563  * and L2TPv3 data frames here.
564  *
565  * L2TPv2 Data Message Header
566  *
567  *  0                   1                   2                   3
568  *  0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
569  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
570  * |T|L|x|x|S|x|O|P|x|x|x|x|  Ver  |          Length (opt)         |
571  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
572  * |           Tunnel ID           |           Session ID          |
573  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
574  * |             Ns (opt)          |             Nr (opt)          |
575  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
576  * |      Offset Size (opt)        |    Offset pad... (opt)
577  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
578  *
579  * Data frames are marked by T=0. All other fields are the same as
580  * those in L2TP control frames.
581  *
582  * L2TPv3 Data Message Header
583  *
584  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
585  * |                      L2TP Session Header                      |
586  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
587  * |                      L2-Specific Sublayer                     |
588  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
589  * |                        Tunnel Payload                      ...
590  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
591  *
592  * L2TPv3 Session Header Over IP
593  *
594  *  0                   1                   2                   3
595  *  0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
596  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
597  * |                           Session ID                          |
598  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
599  * |               Cookie (optional, maximum 64 bits)...
600  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
601  *                                                                 |
602  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
603  *
604  * L2TPv3 L2-Specific Sublayer Format
605  *
606  *  0                   1                   2                   3
607  *  0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
608  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
609  * |x|S|x|x|x|x|x|x|              Sequence Number                  |
610  * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
611  *
612  * Cookie value and sublayer format are negotiated with the peer when
613  * the session is set up. Unlike L2TPv2, we do not need to parse the
614  * packet header to determine if optional fields are present.
615  *
616  * Caller must already have parsed the frame and determined that it is
617  * a data (not control) frame before coming here. Fields up to the
618  * session-id have already been parsed and ptr points to the data
619  * after the session-id.
620  */
621 void l2tp_recv_common(struct l2tp_session *session, struct sk_buff *skb,
622 		      unsigned char *ptr, unsigned char *optr, u16 hdrflags,
623 		      int length, int (*payload_hook)(struct sk_buff *skb))
624 {
625 	struct l2tp_tunnel *tunnel = session->tunnel;
626 	int offset;
627 	u32 ns, nr;
628 
629 	/* Parse and check optional cookie */
630 	if (session->peer_cookie_len > 0) {
631 		if (memcmp(ptr, &session->peer_cookie[0], session->peer_cookie_len)) {
632 			l2tp_info(tunnel, L2TP_MSG_DATA,
633 				  "%s: cookie mismatch (%u/%u). Discarding.\n",
634 				  tunnel->name, tunnel->tunnel_id,
635 				  session->session_id);
636 			atomic_long_inc(&session->stats.rx_cookie_discards);
637 			goto discard;
638 		}
639 		ptr += session->peer_cookie_len;
640 	}
641 
642 	/* Handle the optional sequence numbers. Sequence numbers are
643 	 * in different places for L2TPv2 and L2TPv3.
644 	 *
645 	 * If we are the LAC, enable/disable sequence numbers under
646 	 * the control of the LNS.  If no sequence numbers present but
647 	 * we were expecting them, discard frame.
648 	 */
649 	ns = nr = 0;
650 	L2TP_SKB_CB(skb)->has_seq = 0;
651 	if (tunnel->version == L2TP_HDR_VER_2) {
652 		if (hdrflags & L2TP_HDRFLAG_S) {
653 			ns = ntohs(*(__be16 *) ptr);
654 			ptr += 2;
655 			nr = ntohs(*(__be16 *) ptr);
656 			ptr += 2;
657 
658 			/* Store L2TP info in the skb */
659 			L2TP_SKB_CB(skb)->ns = ns;
660 			L2TP_SKB_CB(skb)->has_seq = 1;
661 
662 			l2tp_dbg(session, L2TP_MSG_SEQ,
663 				 "%s: recv data ns=%u, nr=%u, session nr=%u\n",
664 				 session->name, ns, nr, session->nr);
665 		}
666 	} else if (session->l2specific_type == L2TP_L2SPECTYPE_DEFAULT) {
667 		u32 l2h = ntohl(*(__be32 *) ptr);
668 
669 		if (l2h & 0x40000000) {
670 			ns = l2h & 0x00ffffff;
671 
672 			/* Store L2TP info in the skb */
673 			L2TP_SKB_CB(skb)->ns = ns;
674 			L2TP_SKB_CB(skb)->has_seq = 1;
675 
676 			l2tp_dbg(session, L2TP_MSG_SEQ,
677 				 "%s: recv data ns=%u, session nr=%u\n",
678 				 session->name, ns, session->nr);
679 		}
680 		ptr += 4;
681 	}
682 
683 	if (L2TP_SKB_CB(skb)->has_seq) {
684 		/* Received a packet with sequence numbers. If we're the LNS,
685 		 * check if we sre sending sequence numbers and if not,
686 		 * configure it so.
687 		 */
688 		if ((!session->lns_mode) && (!session->send_seq)) {
689 			l2tp_info(session, L2TP_MSG_SEQ,
690 				  "%s: requested to enable seq numbers by LNS\n",
691 				  session->name);
692 			session->send_seq = 1;
693 			l2tp_session_set_header_len(session, tunnel->version);
694 		}
695 	} else {
696 		/* No sequence numbers.
697 		 * If user has configured mandatory sequence numbers, discard.
698 		 */
699 		if (session->recv_seq) {
700 			l2tp_warn(session, L2TP_MSG_SEQ,
701 				  "%s: recv data has no seq numbers when required. Discarding.\n",
702 				  session->name);
703 			atomic_long_inc(&session->stats.rx_seq_discards);
704 			goto discard;
705 		}
706 
707 		/* If we're the LAC and we're sending sequence numbers, the
708 		 * LNS has requested that we no longer send sequence numbers.
709 		 * If we're the LNS and we're sending sequence numbers, the
710 		 * LAC is broken. Discard the frame.
711 		 */
712 		if ((!session->lns_mode) && (session->send_seq)) {
713 			l2tp_info(session, L2TP_MSG_SEQ,
714 				  "%s: requested to disable seq numbers by LNS\n",
715 				  session->name);
716 			session->send_seq = 0;
717 			l2tp_session_set_header_len(session, tunnel->version);
718 		} else if (session->send_seq) {
719 			l2tp_warn(session, L2TP_MSG_SEQ,
720 				  "%s: recv data has no seq numbers when required. Discarding.\n",
721 				  session->name);
722 			atomic_long_inc(&session->stats.rx_seq_discards);
723 			goto discard;
724 		}
725 	}
726 
727 	/* Session data offset is defined only for L2TPv2 and is
728 	 * indicated by an optional 16-bit value in the header.
729 	 */
730 	if (tunnel->version == L2TP_HDR_VER_2) {
731 		/* If offset bit set, skip it. */
732 		if (hdrflags & L2TP_HDRFLAG_O) {
733 			offset = ntohs(*(__be16 *)ptr);
734 			ptr += 2 + offset;
735 		}
736 	}
737 
738 	offset = ptr - optr;
739 	if (!pskb_may_pull(skb, offset))
740 		goto discard;
741 
742 	__skb_pull(skb, offset);
743 
744 	/* If caller wants to process the payload before we queue the
745 	 * packet, do so now.
746 	 */
747 	if (payload_hook)
748 		if ((*payload_hook)(skb))
749 			goto discard;
750 
751 	/* Prepare skb for adding to the session's reorder_q.  Hold
752 	 * packets for max reorder_timeout or 1 second if not
753 	 * reordering.
754 	 */
755 	L2TP_SKB_CB(skb)->length = length;
756 	L2TP_SKB_CB(skb)->expires = jiffies +
757 		(session->reorder_timeout ? session->reorder_timeout : HZ);
758 
759 	/* Add packet to the session's receive queue. Reordering is done here, if
760 	 * enabled. Saved L2TP protocol info is stored in skb->sb[].
761 	 */
762 	if (L2TP_SKB_CB(skb)->has_seq) {
763 		if (l2tp_recv_data_seq(session, skb))
764 			goto discard;
765 	} else {
766 		/* No sequence numbers. Add the skb to the tail of the
767 		 * reorder queue. This ensures that it will be
768 		 * delivered after all previous sequenced skbs.
769 		 */
770 		skb_queue_tail(&session->reorder_q, skb);
771 	}
772 
773 	/* Try to dequeue as many skbs from reorder_q as we can. */
774 	l2tp_recv_dequeue(session);
775 
776 	return;
777 
778 discard:
779 	atomic_long_inc(&session->stats.rx_errors);
780 	kfree_skb(skb);
781 }
782 EXPORT_SYMBOL(l2tp_recv_common);
783 
784 /* Drop skbs from the session's reorder_q
785  */
786 int l2tp_session_queue_purge(struct l2tp_session *session)
787 {
788 	struct sk_buff *skb = NULL;
789 	BUG_ON(!session);
790 	BUG_ON(session->magic != L2TP_SESSION_MAGIC);
791 	while ((skb = skb_dequeue(&session->reorder_q))) {
792 		atomic_long_inc(&session->stats.rx_errors);
793 		kfree_skb(skb);
794 	}
795 	return 0;
796 }
797 EXPORT_SYMBOL_GPL(l2tp_session_queue_purge);
798 
799 /* Internal UDP receive frame. Do the real work of receiving an L2TP data frame
800  * here. The skb is not on a list when we get here.
801  * Returns 0 if the packet was a data packet and was successfully passed on.
802  * Returns 1 if the packet was not a good data packet and could not be
803  * forwarded.  All such packets are passed up to userspace to deal with.
804  */
805 static int l2tp_udp_recv_core(struct l2tp_tunnel *tunnel, struct sk_buff *skb,
806 			      int (*payload_hook)(struct sk_buff *skb))
807 {
808 	struct l2tp_session *session = NULL;
809 	unsigned char *ptr, *optr;
810 	u16 hdrflags;
811 	u32 tunnel_id, session_id;
812 	u16 version;
813 	int length;
814 
815 	/* UDP has verifed checksum */
816 
817 	/* UDP always verifies the packet length. */
818 	__skb_pull(skb, sizeof(struct udphdr));
819 
820 	/* Short packet? */
821 	if (!pskb_may_pull(skb, L2TP_HDR_SIZE_SEQ)) {
822 		l2tp_info(tunnel, L2TP_MSG_DATA,
823 			  "%s: recv short packet (len=%d)\n",
824 			  tunnel->name, skb->len);
825 		goto error;
826 	}
827 
828 	/* Trace packet contents, if enabled */
829 	if (tunnel->debug & L2TP_MSG_DATA) {
830 		length = min(32u, skb->len);
831 		if (!pskb_may_pull(skb, length))
832 			goto error;
833 
834 		pr_debug("%s: recv\n", tunnel->name);
835 		print_hex_dump_bytes("", DUMP_PREFIX_OFFSET, skb->data, length);
836 	}
837 
838 	/* Point to L2TP header */
839 	optr = ptr = skb->data;
840 
841 	/* Get L2TP header flags */
842 	hdrflags = ntohs(*(__be16 *) ptr);
843 
844 	/* Check protocol version */
845 	version = hdrflags & L2TP_HDR_VER_MASK;
846 	if (version != tunnel->version) {
847 		l2tp_info(tunnel, L2TP_MSG_DATA,
848 			  "%s: recv protocol version mismatch: got %d expected %d\n",
849 			  tunnel->name, version, tunnel->version);
850 		goto error;
851 	}
852 
853 	/* Get length of L2TP packet */
854 	length = skb->len;
855 
856 	/* If type is control packet, it is handled by userspace. */
857 	if (hdrflags & L2TP_HDRFLAG_T) {
858 		l2tp_dbg(tunnel, L2TP_MSG_DATA,
859 			 "%s: recv control packet, len=%d\n",
860 			 tunnel->name, length);
861 		goto error;
862 	}
863 
864 	/* Skip flags */
865 	ptr += 2;
866 
867 	if (tunnel->version == L2TP_HDR_VER_2) {
868 		/* If length is present, skip it */
869 		if (hdrflags & L2TP_HDRFLAG_L)
870 			ptr += 2;
871 
872 		/* Extract tunnel and session ID */
873 		tunnel_id = ntohs(*(__be16 *) ptr);
874 		ptr += 2;
875 		session_id = ntohs(*(__be16 *) ptr);
876 		ptr += 2;
877 	} else {
878 		ptr += 2;	/* skip reserved bits */
879 		tunnel_id = tunnel->tunnel_id;
880 		session_id = ntohl(*(__be32 *) ptr);
881 		ptr += 4;
882 	}
883 
884 	/* Find the session context */
885 	session = l2tp_session_get(tunnel->l2tp_net, tunnel, session_id);
886 	if (!session || !session->recv_skb) {
887 		if (session)
888 			l2tp_session_dec_refcount(session);
889 
890 		/* Not found? Pass to userspace to deal with */
891 		l2tp_info(tunnel, L2TP_MSG_DATA,
892 			  "%s: no session found (%u/%u). Passing up.\n",
893 			  tunnel->name, tunnel_id, session_id);
894 		goto error;
895 	}
896 
897 	l2tp_recv_common(session, skb, ptr, optr, hdrflags, length, payload_hook);
898 	l2tp_session_dec_refcount(session);
899 
900 	return 0;
901 
902 error:
903 	/* Put UDP header back */
904 	__skb_push(skb, sizeof(struct udphdr));
905 
906 	return 1;
907 }
908 
909 /* UDP encapsulation receive handler. See net/ipv4/udp.c.
910  * Return codes:
911  * 0 : success.
912  * <0: error
913  * >0: skb should be passed up to userspace as UDP.
914  */
915 int l2tp_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
916 {
917 	struct l2tp_tunnel *tunnel;
918 
919 	tunnel = l2tp_tunnel(sk);
920 	if (tunnel == NULL)
921 		goto pass_up;
922 
923 	l2tp_dbg(tunnel, L2TP_MSG_DATA, "%s: received %d bytes\n",
924 		 tunnel->name, skb->len);
925 
926 	if (l2tp_udp_recv_core(tunnel, skb, tunnel->recv_payload_hook))
927 		goto pass_up;
928 
929 	return 0;
930 
931 pass_up:
932 	return 1;
933 }
934 EXPORT_SYMBOL_GPL(l2tp_udp_encap_recv);
935 
936 /************************************************************************
937  * Transmit handling
938  ***********************************************************************/
939 
940 /* Build an L2TP header for the session into the buffer provided.
941  */
942 static int l2tp_build_l2tpv2_header(struct l2tp_session *session, void *buf)
943 {
944 	struct l2tp_tunnel *tunnel = session->tunnel;
945 	__be16 *bufp = buf;
946 	__be16 *optr = buf;
947 	u16 flags = L2TP_HDR_VER_2;
948 	u32 tunnel_id = tunnel->peer_tunnel_id;
949 	u32 session_id = session->peer_session_id;
950 
951 	if (session->send_seq)
952 		flags |= L2TP_HDRFLAG_S;
953 
954 	/* Setup L2TP header. */
955 	*bufp++ = htons(flags);
956 	*bufp++ = htons(tunnel_id);
957 	*bufp++ = htons(session_id);
958 	if (session->send_seq) {
959 		*bufp++ = htons(session->ns);
960 		*bufp++ = 0;
961 		session->ns++;
962 		session->ns &= 0xffff;
963 		l2tp_dbg(session, L2TP_MSG_SEQ, "%s: updated ns to %u\n",
964 			 session->name, session->ns);
965 	}
966 
967 	return bufp - optr;
968 }
969 
970 static int l2tp_build_l2tpv3_header(struct l2tp_session *session, void *buf)
971 {
972 	struct l2tp_tunnel *tunnel = session->tunnel;
973 	char *bufp = buf;
974 	char *optr = bufp;
975 
976 	/* Setup L2TP header. The header differs slightly for UDP and
977 	 * IP encapsulations. For UDP, there is 4 bytes of flags.
978 	 */
979 	if (tunnel->encap == L2TP_ENCAPTYPE_UDP) {
980 		u16 flags = L2TP_HDR_VER_3;
981 		*((__be16 *) bufp) = htons(flags);
982 		bufp += 2;
983 		*((__be16 *) bufp) = 0;
984 		bufp += 2;
985 	}
986 
987 	*((__be32 *) bufp) = htonl(session->peer_session_id);
988 	bufp += 4;
989 	if (session->cookie_len) {
990 		memcpy(bufp, &session->cookie[0], session->cookie_len);
991 		bufp += session->cookie_len;
992 	}
993 	if (session->l2specific_type == L2TP_L2SPECTYPE_DEFAULT) {
994 		u32 l2h = 0;
995 
996 		if (session->send_seq) {
997 			l2h = 0x40000000 | session->ns;
998 			session->ns++;
999 			session->ns &= 0xffffff;
1000 			l2tp_dbg(session, L2TP_MSG_SEQ,
1001 				 "%s: updated ns to %u\n",
1002 				 session->name, session->ns);
1003 		}
1004 
1005 		*((__be32 *)bufp) = htonl(l2h);
1006 		bufp += 4;
1007 	}
1008 
1009 	return bufp - optr;
1010 }
1011 
1012 static int l2tp_xmit_core(struct l2tp_session *session, struct sk_buff *skb,
1013 			  struct flowi *fl, size_t data_len)
1014 {
1015 	struct l2tp_tunnel *tunnel = session->tunnel;
1016 	unsigned int len = skb->len;
1017 	int error;
1018 
1019 	/* Debug */
1020 	if (session->send_seq)
1021 		l2tp_dbg(session, L2TP_MSG_DATA, "%s: send %zd bytes, ns=%u\n",
1022 			 session->name, data_len, session->ns - 1);
1023 	else
1024 		l2tp_dbg(session, L2TP_MSG_DATA, "%s: send %zd bytes\n",
1025 			 session->name, data_len);
1026 
1027 	if (session->debug & L2TP_MSG_DATA) {
1028 		int uhlen = (tunnel->encap == L2TP_ENCAPTYPE_UDP) ? sizeof(struct udphdr) : 0;
1029 		unsigned char *datap = skb->data + uhlen;
1030 
1031 		pr_debug("%s: xmit\n", session->name);
1032 		print_hex_dump_bytes("", DUMP_PREFIX_OFFSET,
1033 				     datap, min_t(size_t, 32, len - uhlen));
1034 	}
1035 
1036 	/* Queue the packet to IP for output */
1037 	skb->ignore_df = 1;
1038 #if IS_ENABLED(CONFIG_IPV6)
1039 	if (l2tp_sk_is_v6(tunnel->sock))
1040 		error = inet6_csk_xmit(tunnel->sock, skb, NULL);
1041 	else
1042 #endif
1043 		error = ip_queue_xmit(tunnel->sock, skb, fl);
1044 
1045 	/* Update stats */
1046 	if (error >= 0) {
1047 		atomic_long_inc(&tunnel->stats.tx_packets);
1048 		atomic_long_add(len, &tunnel->stats.tx_bytes);
1049 		atomic_long_inc(&session->stats.tx_packets);
1050 		atomic_long_add(len, &session->stats.tx_bytes);
1051 	} else {
1052 		atomic_long_inc(&tunnel->stats.tx_errors);
1053 		atomic_long_inc(&session->stats.tx_errors);
1054 	}
1055 
1056 	return 0;
1057 }
1058 
1059 /* If caller requires the skb to have a ppp header, the header must be
1060  * inserted in the skb data before calling this function.
1061  */
1062 int l2tp_xmit_skb(struct l2tp_session *session, struct sk_buff *skb, int hdr_len)
1063 {
1064 	int data_len = skb->len;
1065 	struct l2tp_tunnel *tunnel = session->tunnel;
1066 	struct sock *sk = tunnel->sock;
1067 	struct flowi *fl;
1068 	struct udphdr *uh;
1069 	struct inet_sock *inet;
1070 	int headroom;
1071 	int uhlen = (tunnel->encap == L2TP_ENCAPTYPE_UDP) ? sizeof(struct udphdr) : 0;
1072 	int udp_len;
1073 	int ret = NET_XMIT_SUCCESS;
1074 
1075 	/* Check that there's enough headroom in the skb to insert IP,
1076 	 * UDP and L2TP headers. If not enough, expand it to
1077 	 * make room. Adjust truesize.
1078 	 */
1079 	headroom = NET_SKB_PAD + sizeof(struct iphdr) +
1080 		uhlen + hdr_len;
1081 	if (skb_cow_head(skb, headroom)) {
1082 		kfree_skb(skb);
1083 		return NET_XMIT_DROP;
1084 	}
1085 
1086 	/* Setup L2TP header */
1087 	session->build_header(session, __skb_push(skb, hdr_len));
1088 
1089 	/* Reset skb netfilter state */
1090 	memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1091 	IPCB(skb)->flags &= ~(IPSKB_XFRM_TUNNEL_SIZE | IPSKB_XFRM_TRANSFORMED |
1092 			      IPSKB_REROUTED);
1093 	nf_reset(skb);
1094 
1095 	bh_lock_sock(sk);
1096 	if (sock_owned_by_user(sk)) {
1097 		kfree_skb(skb);
1098 		ret = NET_XMIT_DROP;
1099 		goto out_unlock;
1100 	}
1101 
1102 	/* The user-space may change the connection status for the user-space
1103 	 * provided socket at run time: we must check it under the socket lock
1104 	 */
1105 	if (tunnel->fd >= 0 && sk->sk_state != TCP_ESTABLISHED) {
1106 		kfree_skb(skb);
1107 		ret = NET_XMIT_DROP;
1108 		goto out_unlock;
1109 	}
1110 
1111 	/* Get routing info from the tunnel socket */
1112 	skb_dst_drop(skb);
1113 	skb_dst_set(skb, dst_clone(__sk_dst_check(sk, 0)));
1114 
1115 	inet = inet_sk(sk);
1116 	fl = &inet->cork.fl;
1117 	switch (tunnel->encap) {
1118 	case L2TP_ENCAPTYPE_UDP:
1119 		/* Setup UDP header */
1120 		__skb_push(skb, sizeof(*uh));
1121 		skb_reset_transport_header(skb);
1122 		uh = udp_hdr(skb);
1123 		uh->source = inet->inet_sport;
1124 		uh->dest = inet->inet_dport;
1125 		udp_len = uhlen + hdr_len + data_len;
1126 		uh->len = htons(udp_len);
1127 
1128 		/* Calculate UDP checksum if configured to do so */
1129 #if IS_ENABLED(CONFIG_IPV6)
1130 		if (l2tp_sk_is_v6(sk))
1131 			udp6_set_csum(udp_get_no_check6_tx(sk),
1132 				      skb, &inet6_sk(sk)->saddr,
1133 				      &sk->sk_v6_daddr, udp_len);
1134 		else
1135 #endif
1136 		udp_set_csum(sk->sk_no_check_tx, skb, inet->inet_saddr,
1137 			     inet->inet_daddr, udp_len);
1138 		break;
1139 
1140 	case L2TP_ENCAPTYPE_IP:
1141 		break;
1142 	}
1143 
1144 	l2tp_xmit_core(session, skb, fl, data_len);
1145 out_unlock:
1146 	bh_unlock_sock(sk);
1147 
1148 	return ret;
1149 }
1150 EXPORT_SYMBOL_GPL(l2tp_xmit_skb);
1151 
1152 /*****************************************************************************
1153  * Tinnel and session create/destroy.
1154  *****************************************************************************/
1155 
1156 /* Tunnel socket destruct hook.
1157  * The tunnel context is deleted only when all session sockets have been
1158  * closed.
1159  */
1160 static void l2tp_tunnel_destruct(struct sock *sk)
1161 {
1162 	struct l2tp_tunnel *tunnel = l2tp_tunnel(sk);
1163 
1164 	if (tunnel == NULL)
1165 		goto end;
1166 
1167 	l2tp_info(tunnel, L2TP_MSG_CONTROL, "%s: closing...\n", tunnel->name);
1168 
1169 	/* Disable udp encapsulation */
1170 	switch (tunnel->encap) {
1171 	case L2TP_ENCAPTYPE_UDP:
1172 		/* No longer an encapsulation socket. See net/ipv4/udp.c */
1173 		(udp_sk(sk))->encap_type = 0;
1174 		(udp_sk(sk))->encap_rcv = NULL;
1175 		(udp_sk(sk))->encap_destroy = NULL;
1176 		break;
1177 	case L2TP_ENCAPTYPE_IP:
1178 		break;
1179 	}
1180 
1181 	/* Remove hooks into tunnel socket */
1182 	sk->sk_destruct = tunnel->old_sk_destruct;
1183 	sk->sk_user_data = NULL;
1184 
1185 	/* Call the original destructor */
1186 	if (sk->sk_destruct)
1187 		(*sk->sk_destruct)(sk);
1188 
1189 	kfree_rcu(tunnel, rcu);
1190 end:
1191 	return;
1192 }
1193 
1194 /* When the tunnel is closed, all the attached sessions need to go too.
1195  */
1196 void l2tp_tunnel_closeall(struct l2tp_tunnel *tunnel)
1197 {
1198 	int hash;
1199 	struct hlist_node *walk;
1200 	struct hlist_node *tmp;
1201 	struct l2tp_session *session;
1202 
1203 	BUG_ON(tunnel == NULL);
1204 
1205 	l2tp_info(tunnel, L2TP_MSG_CONTROL, "%s: closing all sessions...\n",
1206 		  tunnel->name);
1207 
1208 	write_lock_bh(&tunnel->hlist_lock);
1209 	tunnel->acpt_newsess = false;
1210 	for (hash = 0; hash < L2TP_HASH_SIZE; hash++) {
1211 again:
1212 		hlist_for_each_safe(walk, tmp, &tunnel->session_hlist[hash]) {
1213 			session = hlist_entry(walk, struct l2tp_session, hlist);
1214 
1215 			l2tp_info(session, L2TP_MSG_CONTROL,
1216 				  "%s: closing session\n", session->name);
1217 
1218 			hlist_del_init(&session->hlist);
1219 
1220 			if (test_and_set_bit(0, &session->dead))
1221 				goto again;
1222 
1223 			write_unlock_bh(&tunnel->hlist_lock);
1224 
1225 			__l2tp_session_unhash(session);
1226 			l2tp_session_queue_purge(session);
1227 
1228 			if (session->session_close != NULL)
1229 				(*session->session_close)(session);
1230 
1231 			l2tp_session_dec_refcount(session);
1232 
1233 			write_lock_bh(&tunnel->hlist_lock);
1234 
1235 			/* Now restart from the beginning of this hash
1236 			 * chain.  We always remove a session from the
1237 			 * list so we are guaranteed to make forward
1238 			 * progress.
1239 			 */
1240 			goto again;
1241 		}
1242 	}
1243 	write_unlock_bh(&tunnel->hlist_lock);
1244 }
1245 EXPORT_SYMBOL_GPL(l2tp_tunnel_closeall);
1246 
1247 /* Tunnel socket destroy hook for UDP encapsulation */
1248 static void l2tp_udp_encap_destroy(struct sock *sk)
1249 {
1250 	struct l2tp_tunnel *tunnel = l2tp_tunnel(sk);
1251 
1252 	if (tunnel)
1253 		l2tp_tunnel_delete(tunnel);
1254 }
1255 
1256 /* Workqueue tunnel deletion function */
1257 static void l2tp_tunnel_del_work(struct work_struct *work)
1258 {
1259 	struct l2tp_tunnel *tunnel = container_of(work, struct l2tp_tunnel,
1260 						  del_work);
1261 	struct sock *sk = tunnel->sock;
1262 	struct socket *sock = sk->sk_socket;
1263 	struct l2tp_net *pn;
1264 
1265 	l2tp_tunnel_closeall(tunnel);
1266 
1267 	/* If the tunnel socket was created within the kernel, use
1268 	 * the sk API to release it here.
1269 	 */
1270 	if (tunnel->fd < 0) {
1271 		if (sock) {
1272 			kernel_sock_shutdown(sock, SHUT_RDWR);
1273 			sock_release(sock);
1274 		}
1275 	}
1276 
1277 	/* Remove the tunnel struct from the tunnel list */
1278 	pn = l2tp_pernet(tunnel->l2tp_net);
1279 	spin_lock_bh(&pn->l2tp_tunnel_list_lock);
1280 	list_del_rcu(&tunnel->list);
1281 	spin_unlock_bh(&pn->l2tp_tunnel_list_lock);
1282 
1283 	/* drop initial ref */
1284 	l2tp_tunnel_dec_refcount(tunnel);
1285 
1286 	/* drop workqueue ref */
1287 	l2tp_tunnel_dec_refcount(tunnel);
1288 }
1289 
1290 /* Create a socket for the tunnel, if one isn't set up by
1291  * userspace. This is used for static tunnels where there is no
1292  * managing L2TP daemon.
1293  *
1294  * Since we don't want these sockets to keep a namespace alive by
1295  * themselves, we drop the socket's namespace refcount after creation.
1296  * These sockets are freed when the namespace exits using the pernet
1297  * exit hook.
1298  */
1299 static int l2tp_tunnel_sock_create(struct net *net,
1300 				u32 tunnel_id,
1301 				u32 peer_tunnel_id,
1302 				struct l2tp_tunnel_cfg *cfg,
1303 				struct socket **sockp)
1304 {
1305 	int err = -EINVAL;
1306 	struct socket *sock = NULL;
1307 	struct udp_port_cfg udp_conf;
1308 
1309 	switch (cfg->encap) {
1310 	case L2TP_ENCAPTYPE_UDP:
1311 		memset(&udp_conf, 0, sizeof(udp_conf));
1312 
1313 #if IS_ENABLED(CONFIG_IPV6)
1314 		if (cfg->local_ip6 && cfg->peer_ip6) {
1315 			udp_conf.family = AF_INET6;
1316 			memcpy(&udp_conf.local_ip6, cfg->local_ip6,
1317 			       sizeof(udp_conf.local_ip6));
1318 			memcpy(&udp_conf.peer_ip6, cfg->peer_ip6,
1319 			       sizeof(udp_conf.peer_ip6));
1320 			udp_conf.use_udp6_tx_checksums =
1321 			  ! cfg->udp6_zero_tx_checksums;
1322 			udp_conf.use_udp6_rx_checksums =
1323 			  ! cfg->udp6_zero_rx_checksums;
1324 		} else
1325 #endif
1326 		{
1327 			udp_conf.family = AF_INET;
1328 			udp_conf.local_ip = cfg->local_ip;
1329 			udp_conf.peer_ip = cfg->peer_ip;
1330 			udp_conf.use_udp_checksums = cfg->use_udp_checksums;
1331 		}
1332 
1333 		udp_conf.local_udp_port = htons(cfg->local_udp_port);
1334 		udp_conf.peer_udp_port = htons(cfg->peer_udp_port);
1335 
1336 		err = udp_sock_create(net, &udp_conf, &sock);
1337 		if (err < 0)
1338 			goto out;
1339 
1340 		break;
1341 
1342 	case L2TP_ENCAPTYPE_IP:
1343 #if IS_ENABLED(CONFIG_IPV6)
1344 		if (cfg->local_ip6 && cfg->peer_ip6) {
1345 			struct sockaddr_l2tpip6 ip6_addr = {0};
1346 
1347 			err = sock_create_kern(net, AF_INET6, SOCK_DGRAM,
1348 					  IPPROTO_L2TP, &sock);
1349 			if (err < 0)
1350 				goto out;
1351 
1352 			ip6_addr.l2tp_family = AF_INET6;
1353 			memcpy(&ip6_addr.l2tp_addr, cfg->local_ip6,
1354 			       sizeof(ip6_addr.l2tp_addr));
1355 			ip6_addr.l2tp_conn_id = tunnel_id;
1356 			err = kernel_bind(sock, (struct sockaddr *) &ip6_addr,
1357 					  sizeof(ip6_addr));
1358 			if (err < 0)
1359 				goto out;
1360 
1361 			ip6_addr.l2tp_family = AF_INET6;
1362 			memcpy(&ip6_addr.l2tp_addr, cfg->peer_ip6,
1363 			       sizeof(ip6_addr.l2tp_addr));
1364 			ip6_addr.l2tp_conn_id = peer_tunnel_id;
1365 			err = kernel_connect(sock,
1366 					     (struct sockaddr *) &ip6_addr,
1367 					     sizeof(ip6_addr), 0);
1368 			if (err < 0)
1369 				goto out;
1370 		} else
1371 #endif
1372 		{
1373 			struct sockaddr_l2tpip ip_addr = {0};
1374 
1375 			err = sock_create_kern(net, AF_INET, SOCK_DGRAM,
1376 					  IPPROTO_L2TP, &sock);
1377 			if (err < 0)
1378 				goto out;
1379 
1380 			ip_addr.l2tp_family = AF_INET;
1381 			ip_addr.l2tp_addr = cfg->local_ip;
1382 			ip_addr.l2tp_conn_id = tunnel_id;
1383 			err = kernel_bind(sock, (struct sockaddr *) &ip_addr,
1384 					  sizeof(ip_addr));
1385 			if (err < 0)
1386 				goto out;
1387 
1388 			ip_addr.l2tp_family = AF_INET;
1389 			ip_addr.l2tp_addr = cfg->peer_ip;
1390 			ip_addr.l2tp_conn_id = peer_tunnel_id;
1391 			err = kernel_connect(sock, (struct sockaddr *) &ip_addr,
1392 					     sizeof(ip_addr), 0);
1393 			if (err < 0)
1394 				goto out;
1395 		}
1396 		break;
1397 
1398 	default:
1399 		goto out;
1400 	}
1401 
1402 out:
1403 	*sockp = sock;
1404 	if ((err < 0) && sock) {
1405 		kernel_sock_shutdown(sock, SHUT_RDWR);
1406 		sock_release(sock);
1407 		*sockp = NULL;
1408 	}
1409 
1410 	return err;
1411 }
1412 
1413 static struct lock_class_key l2tp_socket_class;
1414 
1415 int l2tp_tunnel_create(struct net *net, int fd, int version, u32 tunnel_id, u32 peer_tunnel_id, struct l2tp_tunnel_cfg *cfg, struct l2tp_tunnel **tunnelp)
1416 {
1417 	struct l2tp_tunnel *tunnel = NULL;
1418 	int err;
1419 	enum l2tp_encap_type encap = L2TP_ENCAPTYPE_UDP;
1420 
1421 	if (cfg != NULL)
1422 		encap = cfg->encap;
1423 
1424 	tunnel = kzalloc(sizeof(struct l2tp_tunnel), GFP_KERNEL);
1425 	if (tunnel == NULL) {
1426 		err = -ENOMEM;
1427 		goto err;
1428 	}
1429 
1430 	tunnel->version = version;
1431 	tunnel->tunnel_id = tunnel_id;
1432 	tunnel->peer_tunnel_id = peer_tunnel_id;
1433 	tunnel->debug = L2TP_DEFAULT_DEBUG_FLAGS;
1434 
1435 	tunnel->magic = L2TP_TUNNEL_MAGIC;
1436 	sprintf(&tunnel->name[0], "tunl %u", tunnel_id);
1437 	rwlock_init(&tunnel->hlist_lock);
1438 	tunnel->acpt_newsess = true;
1439 
1440 	if (cfg != NULL)
1441 		tunnel->debug = cfg->debug;
1442 
1443 	tunnel->encap = encap;
1444 
1445 	refcount_set(&tunnel->ref_count, 1);
1446 	tunnel->fd = fd;
1447 
1448 	/* Init delete workqueue struct */
1449 	INIT_WORK(&tunnel->del_work, l2tp_tunnel_del_work);
1450 
1451 	INIT_LIST_HEAD(&tunnel->list);
1452 
1453 	err = 0;
1454 err:
1455 	if (tunnelp)
1456 		*tunnelp = tunnel;
1457 
1458 	return err;
1459 }
1460 EXPORT_SYMBOL_GPL(l2tp_tunnel_create);
1461 
1462 static int l2tp_validate_socket(const struct sock *sk, const struct net *net,
1463 				enum l2tp_encap_type encap)
1464 {
1465 	if (!net_eq(sock_net(sk), net))
1466 		return -EINVAL;
1467 
1468 	if (sk->sk_type != SOCK_DGRAM)
1469 		return -EPROTONOSUPPORT;
1470 
1471 	if ((encap == L2TP_ENCAPTYPE_UDP && sk->sk_protocol != IPPROTO_UDP) ||
1472 	    (encap == L2TP_ENCAPTYPE_IP && sk->sk_protocol != IPPROTO_L2TP))
1473 		return -EPROTONOSUPPORT;
1474 
1475 	if (sk->sk_user_data)
1476 		return -EBUSY;
1477 
1478 	return 0;
1479 }
1480 
1481 int l2tp_tunnel_register(struct l2tp_tunnel *tunnel, struct net *net,
1482 			 struct l2tp_tunnel_cfg *cfg)
1483 {
1484 	struct l2tp_tunnel *tunnel_walk;
1485 	struct l2tp_net *pn;
1486 	struct socket *sock;
1487 	struct sock *sk;
1488 	int ret;
1489 
1490 	if (tunnel->fd < 0) {
1491 		ret = l2tp_tunnel_sock_create(net, tunnel->tunnel_id,
1492 					      tunnel->peer_tunnel_id, cfg,
1493 					      &sock);
1494 		if (ret < 0)
1495 			goto err;
1496 	} else {
1497 		sock = sockfd_lookup(tunnel->fd, &ret);
1498 		if (!sock)
1499 			goto err;
1500 
1501 		ret = l2tp_validate_socket(sock->sk, net, tunnel->encap);
1502 		if (ret < 0)
1503 			goto err_sock;
1504 	}
1505 
1506 	sk = sock->sk;
1507 
1508 	sock_hold(sk);
1509 	tunnel->sock = sk;
1510 	tunnel->l2tp_net = net;
1511 
1512 	pn = l2tp_pernet(net);
1513 
1514 	spin_lock_bh(&pn->l2tp_tunnel_list_lock);
1515 	list_for_each_entry(tunnel_walk, &pn->l2tp_tunnel_list, list) {
1516 		if (tunnel_walk->tunnel_id == tunnel->tunnel_id) {
1517 			spin_unlock_bh(&pn->l2tp_tunnel_list_lock);
1518 
1519 			ret = -EEXIST;
1520 			goto err_sock;
1521 		}
1522 	}
1523 	list_add_rcu(&tunnel->list, &pn->l2tp_tunnel_list);
1524 	spin_unlock_bh(&pn->l2tp_tunnel_list_lock);
1525 
1526 	if (tunnel->encap == L2TP_ENCAPTYPE_UDP) {
1527 		struct udp_tunnel_sock_cfg udp_cfg = {
1528 			.sk_user_data = tunnel,
1529 			.encap_type = UDP_ENCAP_L2TPINUDP,
1530 			.encap_rcv = l2tp_udp_encap_recv,
1531 			.encap_destroy = l2tp_udp_encap_destroy,
1532 		};
1533 
1534 		setup_udp_tunnel_sock(net, sock, &udp_cfg);
1535 	} else {
1536 		sk->sk_user_data = tunnel;
1537 	}
1538 
1539 	tunnel->old_sk_destruct = sk->sk_destruct;
1540 	sk->sk_destruct = &l2tp_tunnel_destruct;
1541 	lockdep_set_class_and_name(&sk->sk_lock.slock, &l2tp_socket_class,
1542 				   "l2tp_sock");
1543 	sk->sk_allocation = GFP_ATOMIC;
1544 
1545 	if (tunnel->fd >= 0)
1546 		sockfd_put(sock);
1547 
1548 	return 0;
1549 
1550 err_sock:
1551 	if (tunnel->fd < 0)
1552 		sock_release(sock);
1553 	else
1554 		sockfd_put(sock);
1555 err:
1556 	return ret;
1557 }
1558 EXPORT_SYMBOL_GPL(l2tp_tunnel_register);
1559 
1560 /* This function is used by the netlink TUNNEL_DELETE command.
1561  */
1562 void l2tp_tunnel_delete(struct l2tp_tunnel *tunnel)
1563 {
1564 	if (!test_and_set_bit(0, &tunnel->dead)) {
1565 		l2tp_tunnel_inc_refcount(tunnel);
1566 		queue_work(l2tp_wq, &tunnel->del_work);
1567 	}
1568 }
1569 EXPORT_SYMBOL_GPL(l2tp_tunnel_delete);
1570 
1571 /* Really kill the session.
1572  */
1573 void l2tp_session_free(struct l2tp_session *session)
1574 {
1575 	struct l2tp_tunnel *tunnel = session->tunnel;
1576 
1577 	BUG_ON(refcount_read(&session->ref_count) != 0);
1578 
1579 	if (tunnel) {
1580 		BUG_ON(tunnel->magic != L2TP_TUNNEL_MAGIC);
1581 		l2tp_tunnel_dec_refcount(tunnel);
1582 	}
1583 
1584 	kfree(session);
1585 }
1586 EXPORT_SYMBOL_GPL(l2tp_session_free);
1587 
1588 /* Remove an l2tp session from l2tp_core's hash lists.
1589  * Provides a tidyup interface for pseudowire code which can't just route all
1590  * shutdown via. l2tp_session_delete and a pseudowire-specific session_close
1591  * callback.
1592  */
1593 void __l2tp_session_unhash(struct l2tp_session *session)
1594 {
1595 	struct l2tp_tunnel *tunnel = session->tunnel;
1596 
1597 	/* Remove the session from core hashes */
1598 	if (tunnel) {
1599 		/* Remove from the per-tunnel hash */
1600 		write_lock_bh(&tunnel->hlist_lock);
1601 		hlist_del_init(&session->hlist);
1602 		write_unlock_bh(&tunnel->hlist_lock);
1603 
1604 		/* For L2TPv3 we have a per-net hash: remove from there, too */
1605 		if (tunnel->version != L2TP_HDR_VER_2) {
1606 			struct l2tp_net *pn = l2tp_pernet(tunnel->l2tp_net);
1607 			spin_lock_bh(&pn->l2tp_session_hlist_lock);
1608 			hlist_del_init_rcu(&session->global_hlist);
1609 			spin_unlock_bh(&pn->l2tp_session_hlist_lock);
1610 			synchronize_rcu();
1611 		}
1612 	}
1613 }
1614 EXPORT_SYMBOL_GPL(__l2tp_session_unhash);
1615 
1616 /* This function is used by the netlink SESSION_DELETE command and by
1617    pseudowire modules.
1618  */
1619 int l2tp_session_delete(struct l2tp_session *session)
1620 {
1621 	if (test_and_set_bit(0, &session->dead))
1622 		return 0;
1623 
1624 	__l2tp_session_unhash(session);
1625 	l2tp_session_queue_purge(session);
1626 	if (session->session_close != NULL)
1627 		(*session->session_close)(session);
1628 
1629 	l2tp_session_dec_refcount(session);
1630 
1631 	return 0;
1632 }
1633 EXPORT_SYMBOL_GPL(l2tp_session_delete);
1634 
1635 /* We come here whenever a session's send_seq, cookie_len or
1636  * l2specific_type parameters are set.
1637  */
1638 void l2tp_session_set_header_len(struct l2tp_session *session, int version)
1639 {
1640 	if (version == L2TP_HDR_VER_2) {
1641 		session->hdr_len = 6;
1642 		if (session->send_seq)
1643 			session->hdr_len += 4;
1644 	} else {
1645 		session->hdr_len = 4 + session->cookie_len;
1646 		session->hdr_len += l2tp_get_l2specific_len(session);
1647 		if (session->tunnel->encap == L2TP_ENCAPTYPE_UDP)
1648 			session->hdr_len += 4;
1649 	}
1650 
1651 }
1652 EXPORT_SYMBOL_GPL(l2tp_session_set_header_len);
1653 
1654 struct l2tp_session *l2tp_session_create(int priv_size, struct l2tp_tunnel *tunnel, u32 session_id, u32 peer_session_id, struct l2tp_session_cfg *cfg)
1655 {
1656 	struct l2tp_session *session;
1657 
1658 	session = kzalloc(sizeof(struct l2tp_session) + priv_size, GFP_KERNEL);
1659 	if (session != NULL) {
1660 		session->magic = L2TP_SESSION_MAGIC;
1661 		session->tunnel = tunnel;
1662 
1663 		session->session_id = session_id;
1664 		session->peer_session_id = peer_session_id;
1665 		session->nr = 0;
1666 		if (tunnel->version == L2TP_HDR_VER_2)
1667 			session->nr_max = 0xffff;
1668 		else
1669 			session->nr_max = 0xffffff;
1670 		session->nr_window_size = session->nr_max / 2;
1671 		session->nr_oos_count_max = 4;
1672 
1673 		/* Use NR of first received packet */
1674 		session->reorder_skip = 1;
1675 
1676 		sprintf(&session->name[0], "sess %u/%u",
1677 			tunnel->tunnel_id, session->session_id);
1678 
1679 		skb_queue_head_init(&session->reorder_q);
1680 
1681 		INIT_HLIST_NODE(&session->hlist);
1682 		INIT_HLIST_NODE(&session->global_hlist);
1683 
1684 		/* Inherit debug options from tunnel */
1685 		session->debug = tunnel->debug;
1686 
1687 		if (cfg) {
1688 			session->pwtype = cfg->pw_type;
1689 			session->debug = cfg->debug;
1690 			session->mtu = cfg->mtu;
1691 			session->mru = cfg->mru;
1692 			session->send_seq = cfg->send_seq;
1693 			session->recv_seq = cfg->recv_seq;
1694 			session->lns_mode = cfg->lns_mode;
1695 			session->reorder_timeout = cfg->reorder_timeout;
1696 			session->l2specific_type = cfg->l2specific_type;
1697 			session->cookie_len = cfg->cookie_len;
1698 			memcpy(&session->cookie[0], &cfg->cookie[0], cfg->cookie_len);
1699 			session->peer_cookie_len = cfg->peer_cookie_len;
1700 			memcpy(&session->peer_cookie[0], &cfg->peer_cookie[0], cfg->peer_cookie_len);
1701 		}
1702 
1703 		if (tunnel->version == L2TP_HDR_VER_2)
1704 			session->build_header = l2tp_build_l2tpv2_header;
1705 		else
1706 			session->build_header = l2tp_build_l2tpv3_header;
1707 
1708 		l2tp_session_set_header_len(session, tunnel->version);
1709 
1710 		refcount_set(&session->ref_count, 1);
1711 
1712 		return session;
1713 	}
1714 
1715 	return ERR_PTR(-ENOMEM);
1716 }
1717 EXPORT_SYMBOL_GPL(l2tp_session_create);
1718 
1719 /*****************************************************************************
1720  * Init and cleanup
1721  *****************************************************************************/
1722 
1723 static __net_init int l2tp_init_net(struct net *net)
1724 {
1725 	struct l2tp_net *pn = net_generic(net, l2tp_net_id);
1726 	int hash;
1727 
1728 	INIT_LIST_HEAD(&pn->l2tp_tunnel_list);
1729 	spin_lock_init(&pn->l2tp_tunnel_list_lock);
1730 
1731 	for (hash = 0; hash < L2TP_HASH_SIZE_2; hash++)
1732 		INIT_HLIST_HEAD(&pn->l2tp_session_hlist[hash]);
1733 
1734 	spin_lock_init(&pn->l2tp_session_hlist_lock);
1735 
1736 	return 0;
1737 }
1738 
1739 static __net_exit void l2tp_exit_net(struct net *net)
1740 {
1741 	struct l2tp_net *pn = l2tp_pernet(net);
1742 	struct l2tp_tunnel *tunnel = NULL;
1743 	int hash;
1744 
1745 	rcu_read_lock_bh();
1746 	list_for_each_entry_rcu(tunnel, &pn->l2tp_tunnel_list, list) {
1747 		l2tp_tunnel_delete(tunnel);
1748 	}
1749 	rcu_read_unlock_bh();
1750 
1751 	flush_workqueue(l2tp_wq);
1752 	rcu_barrier();
1753 
1754 	for (hash = 0; hash < L2TP_HASH_SIZE_2; hash++)
1755 		WARN_ON_ONCE(!hlist_empty(&pn->l2tp_session_hlist[hash]));
1756 }
1757 
1758 static struct pernet_operations l2tp_net_ops = {
1759 	.init = l2tp_init_net,
1760 	.exit = l2tp_exit_net,
1761 	.id   = &l2tp_net_id,
1762 	.size = sizeof(struct l2tp_net),
1763 };
1764 
1765 static int __init l2tp_init(void)
1766 {
1767 	int rc = 0;
1768 
1769 	rc = register_pernet_device(&l2tp_net_ops);
1770 	if (rc)
1771 		goto out;
1772 
1773 	l2tp_wq = alloc_workqueue("l2tp", WQ_UNBOUND, 0);
1774 	if (!l2tp_wq) {
1775 		pr_err("alloc_workqueue failed\n");
1776 		unregister_pernet_device(&l2tp_net_ops);
1777 		rc = -ENOMEM;
1778 		goto out;
1779 	}
1780 
1781 	pr_info("L2TP core driver, %s\n", L2TP_DRV_VERSION);
1782 
1783 out:
1784 	return rc;
1785 }
1786 
1787 static void __exit l2tp_exit(void)
1788 {
1789 	unregister_pernet_device(&l2tp_net_ops);
1790 	if (l2tp_wq) {
1791 		destroy_workqueue(l2tp_wq);
1792 		l2tp_wq = NULL;
1793 	}
1794 }
1795 
1796 module_init(l2tp_init);
1797 module_exit(l2tp_exit);
1798 
1799 MODULE_AUTHOR("James Chapman <jchapman@katalix.com>");
1800 MODULE_DESCRIPTION("L2TP core");
1801 MODULE_LICENSE("GPL");
1802 MODULE_VERSION(L2TP_DRV_VERSION);
1803 
1804